Ask HN: Relaunch my website in Ruby/Python or continue with PHP?
Hey All,
My site http://LetsLunch.com is written in PHP - YII framework - we are looking to relaunch it with a new angle/pivot. Any ideas on should we continue to build on existing PHP or switch to more modern languages like ruby/rails, angular etc.. is switching over and rewriting worth it? First off, most of those languages/frameworks aren't even modern, let alone "more modern" than PHP. Next most of them are not significantly faster than PHP, if they're not definitely slower than PHP. If you're not using modern PHP, you can upgrade to PHP7.x faster than a full rewrite. I would avoid rewriting anything non-trivial, because you have no guarantee that your new codebase will be completed in a predictable amount of time, let alone be faster. If a specific sub-module needs to be in a faster language, rewrite that small part in something like golang or elixir. You need to be really careful with full rewrites, you're more likely to end up with a whole year of lost productivity than a success story. I'd recommend learning how to profile and optimize performance _after_ migrating to PHP7+. This is really helpful. I was being pushed to upgrade by the programmers i am trying to hire who say old website code is kind of spaghetti..and a new clean rewrite in node/ruby will go a long way in hiring good bayarea programmers. Stay the course and tune out the fashionistas. We maintain large codebases in all three, and I can only remember one case where any of the languages got in the way (PHP's poor threading support). And even that case wasn't a show stopper--we just used fork() instead. Great business idea BTW. If I lived in a larger city I would definitely try it out! Thanks for the helpful tips. Its hard to live in SF and not feel like missing out on Ruby/rail, node.js and other new fast and easy programming languages. I'd say, if the site can be rewritten quickly, use the language you're most productive in. If the site is small and can be ported in < 1 week might be a good rule. But if your most productive language is PHP, why not leave it? No reason to use Rails, Node, or Python for this. i thought new languages are easier,scalable, more libraries etc since they are new.
But it will take more than a week for sure to port the site data to new rails/node.
Thats why wanted to make sure its well worth switching. I wouldn't say Python or Rails is more scalable, but the libraries may make certain things easier. What libraries do you feel you're missing out on? Rewrite it in Go with Caddy for the backend and plain JS for the front. Check out Laravel if you're sticking with PHP but interested in a new framework. Definitely try out Laravel. I'm not involved in the PHP community, but had a chance to take off a client project and decided to go with Laravel. It was real pleasure working with in combination with writing modern PHP overall. Jeffrey Way's Laracast was of help for me and is worth checking out. As a laravel dev by day (and sometimes night) --- can say Vue.js and Laravel are VERY modern frameworks and constantly updated--Vue is optional and can be swapped for React / Angular or even jquery -- but laravel is a great backend or api for a service oriented architecture. Do you have some ugly corners in your code base? Lots of legacy stuff that you want to rework anyway, regardless the language? Then it might make sense. website is kind of slow and some bugs..so i am wondering if refactoring or bugs fixing is worth it or switching to a new modern language. If your problem is speed, Ruby & Python probably aren't going to help. Query optimizations typically yield a lot of bang for the buck. Once you've harvested all of the low-hanging fruit there, in-memory caching (memcached, redis) can also bring a lot to the table. Ruby and Python could help -- but depends what you're doing. They aren't going to magically make your DB queries faster... but certain libraries can perform many, many times faster. Though, that's the library, not the language. Those languages won't be faster, they're pretty much the same speed. The labor cost of the rewrite would completely dwarf any possible micro-gains from switching between 2 languages that are already nearly the same speed. PHP can be faster than Python, but I'm not aware of anything existing in the PHP space like NumPy. It doesn't sound like OP is working on something that needs NumPy, and if he did, I don't see why he couldn't write a small script in Python to be called by the existing codebase. I completely agree! it also helps to find more programmers here in the bayarea if i switch to ruby, node etc all the cool programming dudes that i want to hire -are not in php. Not many devs want to work on PHP codebases, that's absolutely correct. PHP does not really look great on the resume, unfortunately. But you should be hiring to solve a problem -- seemingly speed and reliability, in your case -- not for a language. But yeah, tell them what stack they will be working on ;) If you were almost anywhere else, you'd have a much easier time finding .NET developers, but then there's the whole licensing can of worms (and .NET core isn't really a viable alternative at the moment). I did PHP before Python. Done CakePHP, Drupal back in the day. Then did some idiorm + paris w/ slim. That said, I'd like to try to sell you on Python. First off, let's begin by saying PHP is a solid web language with mature tools. I haven't looked at the ecosystem in maybe 5 years. Due to this, I can't speak for PHP 7 and the latest and greatest best practices and libraries. But as of 2017, I think python has surpassed PHP in web development. This is due to some fundamental things being baked into the language, and external factors like luck, timing, and momentum. 0. PHP's "lost years": PHP didn't capitalize on standardization, packaging and libraries for a long time. I don't know what the reason was, but it was PHP 5 for a long time and the language didn't "move". This allowed other languages like ruby to gain ground back in 2005-2012. PHP 6 never happened. The community had to pull together and start PHP-FIG and stuff. During those years, a lot happened, other language ecosystems gained significant traction. Sure, there were great strides in individual projects like Drupal and Wordpress. But these CMS and frameworks lacked a package manager during that time. And despite being OOP, developers weren't sharing each other's code the same way Ruby and Python did. 1. Code standardization: Python nails this. PEP8 (code style), PEP257 (code documentation). 1.1 Code aesthetic: While the lack of subcolons and block-style struck me as a gimmick for a while (remember I did PHP before python), I came to find it makes code a hell of a lot easier to read. 2. Package management - People started publishing packages early on. Even if there was distutils/setuptools warts for a couple years, the practice of having a real market for code was great. PHP didn't have that for a long time.. anyone remember PEAR? 3. OOP: Python's OOP makes it very clear to move up the class tree. This sort of winds in with the the code style too, you can easily navigate and traverse large codebases and wrap your brain around chunks of code really fast in python. 4. Imports: Better implemented in python. Much more straight forward and clear. You can pull in a full module from your environment's current packages or relative location (import flask or import . mymodule). You can also import objects like constants, functions and classes scoped in a module, like "from mymodule import myfunction", also import as an alias, "from mymodule import myfunction as a". 5. Community: At this point, Python's community is more active. Though PHP seems to be strong as well. People are building websites in PHP and are still successful at it. That says a lot. 6. Use Domains: Python is being used in big data (pyspark), general data science with pandas, numpy, juypter. The web with django, flask. System utilities like supervisor. People can write mobile games with kivy. There are C speedups like Cython. ORM's like Peewee and SQLAlchemy are mature and superbly documented. 6.1 Web: Python's most popular web framework, Django, is very mature. The ORM is very nice. Migrations. Documentation. Flask, a popular framework for Python, also have excellent documentation and many users. And aside from that, there are many other web libraries like Pyramid, and ones I can't even think off the top of my head. Pyramid is actually one I hear great stuff about haven't even tried yet. 7. Runtime implementations: CPython (default), pypy (python-hosted), Jython for Java and IronPython for .NET 8. Performance: pypy, numpy/pandas has tabular / typed speedups. I'm probably forgetting a few. 9. C/C++ integration: Boost.Python, pybind11 (new and highly cool), CFFI, Cython, Swig, and the CPython's own implementation of C API's, which CPython itself builds upon is at your disposal, and ctypes. 10. Documentation: Python itself, as well as the top libraries in the open source community all use Sphinx to generate documentation. Builds nice static, themed HTML sites with a nifty feature called "intersphinx" to link across project documentation. And it's used at scale in the community. Big thanks to ReadTheDocs.org for providing a place for the community to host. In most cases, the above "nice things" about python have been around in production and used at scale long before I did Python full time. This gives python a significant advantage. Some additional free resources on Python: Thanks for the detailed rundown. How long do you think it will take to rebuild the existing letslunch.com website from scatch in python? just a rough estimate without going for the nitty gritty of table sizes, length, number of lines of code.. letslunch has data of 90,000 users, 60,000 lunches, 12-15 dynamic pages and about 6 api's that we use from linkedin, facebook etc Build it in parallel with your current codebase. If you're not doing so already, create a staging database you can use for developing this. That way if something goes berserk production data stays safe. Get a jumpstart by using Django and creating models that match your database via inspectdb [1]. Most important thing is getting the data to fit and relate as you'd expect. Some recommended plugins are django-debug-toolbar (https://github.com/jazzband/django-debug-toolbar) and django-extensions (https://github.com/django-extensions/django-extensions). Then for Social API logins use https://python-social-auth-docs.readthedocs.io. For Social API's themselves there are python wrappers, like https://github.com/mobolic/facebook-sdk. In the worst case, you'll be querying API's directly with requests (http://docs.python-requests.org). Next, you begin to build views (and their templates) around the models you've created. You're going to want them to run at a parity with what you have in PHP. Use urls.py regular expressions to match your URL's too. This is how you mitigate the risk of doing a "migration". And also a great way to get some experience under the belt while you do it. You should be able to get your site working off a staging database in a read-only way when you do this. Keep the PHP site in production until your ready to switch. If you use JSON/REST, also check out Django REST Framework. http://www.django-rest-framework.org/ I'm going to be making a detailed blog post on django in a few days as well. [1] https://docs.djangoproject.com/en/1.11/howto/legacy-database... Based on the website and code base i have, i just want a rough estimate..are we talking 2 weeks, 2 months or 6 months here to rebuild in a new language? does it provide any advantage that the code was already written and api's work etc in the other language? trying to see if something can be reused or is it gonna be a total build from scatch new website.. > does it provide any advantage that the code was already written and api's work etc in the other language If you build your models to be an exact fit with your current data, yes. > Based on the website and code base i have, i just want a rough estimate..are we talking 2 weeks, 2 months or 6 months here to rebuild in a new language? That's like asking a consultant for a time estimate when they don't know all the details. Factors include: - How much python you know already - How long does it take for you to grasp the fundamentals of virtual environments, packages, importing, etc. > trying to see if something can be reused or is it gonna be a total build from scatch new website.. - Good point: If you end up keeping old API's running on the PHP side and just migrating the HTML stuff first, or only. That is a shortcut to get going faster. - How much templating you have to do. But I think you can do that in maybe (2 Weeks). Django templates are easy to learn and you can extend from a common base. - Database models. I think you should try the inspectdb thing ASAP as a trial to see how easy it is to display any data from your current schema. Be sure to use a copy (or sample) of your prod data in a separate staging database though. You could do that tonight. Getting everything (relations and all) to map perfectly (ORM -> your current db schema) could take a couple of weeks to get perfect. Use this 'managed = False': https://docs.djangoproject.com/en/1.11/howto/legacy-database... - Development hosting: I recommend trying out Heroku. The deployment is straight forward. But outside of that you're going to have to wrap your brain around how to deploy it. nginx + uwsgi + supervisor is a common combo if you're doing it right on the server. - Business logic: Most of this is done inside your views. This could take a month or so. Since django is a big framework and these things work in tandem, it could be 3-6 months. Remember to build it in parallel so the PHP site can continue to be maintained if needed. And create a staging database that's based off your production schema with some sample data in there. Also while you do this, you're going to continually get faster as you gain mastery with python. For a speed boost, you can continue to keep API's running in PHP while you convert other stuff over first. wow..excellent advice. You made me day and week!!